Optimization Problems in Multisensor and Multitarget Tracking

Book excerpt: The objective of this research program is to develop optimization algorithms that solve key problems in multiple target tracking and sensor data fusion. The central problem in multiple target tracking is the data association problem of partitioning sensor reports into tracks and false alarms. New classes of data association problems have been formulated and initial algorithms developed to address cluster tracking, merged measurements, and even sensor resource management in the form of "group-assignments." In a different direction, an efficient k-best algorithm has been developed to approximate the uncertainty in data association, which is ontical for discrimination or combat identification. Statistical Monte Carlo methods are also applicable and are still under investigation. Bias estimation algorithms using known data association such as truth objects and targets of opportunity have been developed. Bias estimation in which data association is unknown is difficult due to the nonconvex and mixed integer nature of the mathematical formulation. Exact and approximate algorithms have been developed and successfully applied to system tracking. As a prerequisite to the development of multiple target tracking approaches to space surveillance, consistent measures of uncertainty for initial orbit determination and the propagation of the uncertainty over time have been developed.